[v3,0/4] x86/time: calibration rendezvous adjustments

Message

Jan Beulich Feb. 9, 2021, 12:53 p.m. UTC

The middle two patches are meant to address a regression reported on
the list under "Problems with APIC on versions 4.9 and later (4.8
works)". In the course of analyzing output from a debugging patch I
ran into another anomaly again, which I thought I should finally try
to address. Hence patch 1. Patch 4 is new in v3 and RFC for now.

While looking closely at the corresponding debugging patch'es output I
noticed a suspicious drift between local and master stime: Measured not
very precisely, local was behind master by about 200ms in about half an
hour. Interestingly the recording of ->master_stime (and hence the not
really inexpensive invocation of read_platform_stime()) looks to be
pretty pointless when CONSTANT_TSC - I haven't been able to spot an
actual consumer. IOW the drift may not be a problem, and we might be
able to eliminate the platform timer reads. (When !CONSTANT_TSC, such
drift would get corrected anyway, by local_time_calibration().)

1: change initiation of the calibration timer
2: adjust time recording time_calibration_tsc_rendezvous()
3: don't move TSC backwards in time_calibration_tsc_rendezvous()
4: re-arrange struct calibration_rendezvous

Jan

Comments

Jan Beulich Feb. 9, 2021, 1:02 p.m. UTC | #1

On 09.02.2021 13:53, Jan Beulich wrote:
> The middle two patches are meant to address a regression reported on
> the list under "Problems with APIC on versions 4.9 and later (4.8
> works)". In the course of analyzing output from a debugging patch I
> ran into another anomaly again, which I thought I should finally try
> to address. Hence patch 1. Patch 4 is new in v3 and RFC for now.

Of course this is the kind of change I'd prefer doing early in a
release cycle. I don't think there are severe risks from patch 1, but
I'm not going to claim patches 2 and 3 are risk free. They fix booting
Xen on a system left in rather awkward state by the firmware. And
they shouldn't affect well behaved modern systems at all (due to
those using a different rendezvous function). While we've been having
this issue for years, I also consider this set a backporting
candidate. Hence I can see reasons pro and con inclusion in 4.15.

Jan

> 1: change initiation of the calibration timer
> 2: adjust time recording time_calibration_tsc_rendezvous()
> 3: don't move TSC backwards in time_calibration_tsc_rendezvous()
> 4: re-arrange struct calibration_rendezvous
> 
> Jan
>

Jan Beulich Feb. 17, 2021, 8:27 a.m. UTC | #2

Ian?

On 09.02.2021 14:02, Jan Beulich wrote:
> On 09.02.2021 13:53, Jan Beulich wrote:
>> The middle two patches are meant to address a regression reported on
>> the list under "Problems with APIC on versions 4.9 and later (4.8
>> works)". In the course of analyzing output from a debugging patch I
>> ran into another anomaly again, which I thought I should finally try
>> to address. Hence patch 1. Patch 4 is new in v3 and RFC for now.
> 
> Of course this is the kind of change I'd prefer doing early in a
> release cycle. I don't think there are severe risks from patch 1, but
> I'm not going to claim patches 2 and 3 are risk free. They fix booting
> Xen on a system left in rather awkward state by the firmware. And
> they shouldn't affect well behaved modern systems at all (due to
> those using a different rendezvous function). While we've been having
> this issue for years, I also consider this set a backporting
> candidate. Hence I can see reasons pro and con inclusion in 4.15.
> 
> Jan
> 
>> 1: change initiation of the calibration timer
>> 2: adjust time recording time_calibration_tsc_rendezvous()
>> 3: don't move TSC backwards in time_calibration_tsc_rendezvous()
>> 4: re-arrange struct calibration_rendezvous
>>
>> Jan
>>
>

Ian Jackson Feb. 19, 2021, 4:06 p.m. UTC | #3

Jan Beulich writes ("[PATCH v3 0/4] x86/time: calibration rendezvous adjustments"):
> The middle two patches are meant to address a regression reported on
> the list under "Problems with APIC on versions 4.9 and later (4.8
> works)". In the course of analyzing output from a debugging patch I
> ran into another anomaly again, which I thought I should finally try
> to address. Hence patch 1. Patch 4 is new in v3 and RFC for now.

For patches 1-3:

Release-Acked-by: Ian Jackson <[e-mail]>

Ian.